Industry Hall Of Fame Introduction Paul Allen Programming Pioneer Tim Berners-Lee Developer Of The World Wide Web Dan Bricklin Creator Of The Electronic Spreadsheet Vint Cerf The Father Of The Internet Ross Cooley Compaq's Channel Champion Larry Ellison Database Dynamo Bronson Ingram King Of Global Distribution Empire Charles Wang Software Mangement Mogul John Warnock Wizard Of Type Steve Wozniak Apple's Engineering Genius Development Teams Introduction
|
The processor that
brought the PC industry to the next level Its beginning was anything but glamorous. In
fact, as Pat Gelsinger recalls, when he went to work on Intel Corp.'s 386 chip
in 1982 company executives had very low expectations for the processor. Instead,
these executives were eagerly awaiting the work from their elite team of
architects, who were working on a glitzier chip in Oregon.
"Sort of the broad view at Intel was that it [the 386] was sort of a transition
project for the real project that was going on in Oregon at the time, which was
going to be the much bigger and exciting solution for the world," said
Gelsinger, now corporate vice president and general manager of Intel's desktop
products group. "This was the gap filler."
But this "gap filler" turned out to be a huge success for Santa Clara,
Calif.-based Intel, and it forever altered the company and the PC industry.
"The 386 is enormously important because it basically established the software
architecture that the entire industry is still based upon today," said Michael
Slater, principal analyst at MicroDesign Resources, Sebastopol, Calif. "It got
us out of the ugliness of the 286 architecture, which was really a
catastrophe."
The original concept of the 386 chip was to devise a 32-bit architectural
extension to Intel's 8086 and 286 microprocessors that would remain compatible
with all the old software, said John Crawford, lead architect on the project.
That concept, developed at the end of 1981, stayed intact at the project's high
conceptual level, he said.
Crawford, now an Intel Fellow and director of microprocessor architecture, came
to the 386 project as a software developer. Gelsinger, meanwhile, was a
technician in quality assurance recruited by the design team for his skill in
running Unix systems.
"So between the leftover architect and the Unix technician, we had pretty humble
beginnings for the project," Gelsinger said.
Plus, the team was pitted against Intel's star-studded team in Oregon. Early on,
Gelsinger's manager handed him "The Soul Of A New Machine," a book that told the
remarkable story of competing design teams at Data General Corp. Gelsinger
remembers his boss telling him: "You're living it."
The 386 team started off with eight people, but its membership grew to nearly
100 by the end of the design phase in July 1985. Despite its lowly start,
expectations for the chip continued to grow over the course of the project.
Gelsinger remembers the end of the 386 design process as very intense.
"That finishing of the chip, when you try to pull the whole design together, is
a really intensive phase where just one little error anywhere in the chip and
you have to rerun the whole thing. You just have to get the whole thing
perfect," Gelsinger said. "We called it giving birth to an elephant."
No one on the chip's development team got much sleep during the elephant's
birth, he said.
"As you're trying to get the chip out the door and in the early testing phase,
you can literally work 24 hours a day, seven days a week," Gelsinger said. "It's
just that much work to do."
The nonstop work created "chip widows" and put a strain on people's private
lives, Crawford said.
Crawford's wife spent a lot of time with Gelsinger's wife while the two men
worked on the new technology and "saw far too much of each other" at Intel,
Gelsinger said.
But that did not dampen spirits. In the drive to finish the development of the
chip, some on the team came up with creative names for three fab teams that
raced to produce the first 386 silicon: Thunderbuns, Hell Rider and Lucky
Runner.
"Thunderbuns was the front-runner for a while and Hell Rider faltered as well,"
Crawford said. "Lucky Runner, with its less-exotic name, finished the race."
Gelsinger, for his part, participated in a practical joke when Lucky Runner
finished its work. He and an engineer on the team, Chip Krasuskopf,
choreographed the stunt after driving to pick up the wafers at Intel's fab in
Folsom, Gelsinger said.
"All the way up, we're playing old Rolling Stones and other rock and roll with
the top down, these two young kids going to get the company jewels," Gelsinger
said.
They returned to Intel headquarters, where the design team was eager to start
testing the chips.
"I come into the room with a box of dead wafers that were all bad, not the real
ones. And I drop them on the floor, trip and fall on top of them, breaking the
wafers up," Gelsinger said.
"There was a gasp that fell across the room because everyone thought we had
destroyed the first 386 wafers," he said.
As Gelsinger embellished his role, pretending to avoid stepping on the wafers
but smashing them even more, one of the managers nervously kneeled down on the
floor, he said.
"He starts picking up pieces, looking for a piece big enough that we can go
test," Gelsinger said. "And then behind me comes Chip with the real wafers."
But when the chip was released into the marketplace in October 1985, it was
certainly no joke.
"We had a wonderful launch and real good acceptance for it," Crawford recalled.
"It was just a couple years later when all the PCs were based on 386s and a lot
of software coming out required 32-bit computing."
The first ad for the 386 PC from Compaq Computer Corp. was the "proudest day of
my professional career," Gelsinger said. The most critical aspect of the 386's
marketing success was Compaq introducing the first 386 machine, he said.
"At the time, IBM [Corp.] had pretty much decided that 32 bits was going to be
the realm of their minicomputers and 16 bits for the 286 family was going to be
sufficient for personal computers," he said.
"The industry was waiting on IBM to introduce its first 386 machine. And it was
not until Compaq introduced its first 386 machine and essentially took a
leadership position in the industry that they forced IBM to respond and bring
out 386--and the entire clone and PC industry was never the same again,"
Gelsinger said.
Many on the 386 design team still work at Intel.
Some, like Gelsinger, have moved out of design. Others, including Crawford,
continue to thrive being in the thick of Intel's core processor development.
Crawford heads the company's joint 64-bit Merced chip project with
Hewlett-Packard Co.
"Intel was able to evolve the 286 design into the 386, adding 32-bit capability,
basically breaking the physical and virtual memory address limits of the 286 in
such a way that it was able to preserve the advantages and compatibility of the
8086 and 286," said Nathan Brookwood, an analyst at Insight 64 in Saratoga,
Calif.
This kept all of the customers happy "because they didn't have to buy new
software and at the same time it opened up the PC architecture for increased
levels of performance and much larger physical memory configuration, all of
which were important," Brookwood said.
|